Structural Damage Identification Based On Modal Sensitivity Analysis

Structures suffer various damages in lifetime and those hard to be detected may develop persistently and destroy the whole structures, thus study on structural damage identification has become a hot topic. Modal parameters are used most frequently among all these methods. Being funded by "Structural damage detection based on piezoelectric information and spectral finite element model updating"(50378041) of National Natural Science Fund of China, a method combining modal sensitivity theory with genetic algorithms has been developed to identify damages of shearing-type multi-storey buildings in this thesis. Modal sensitivity coefficients (those of frequency, modal shape, the slope and the curvature of modal shape) of shearing-type buildings are gained from the wave propagation theory, and the laws and relationships of them are introduced. Further, damage identification is transferred into an optimization problem through sensitivity equations and solved by genetic algorithms. Different damages on 10-storey, 20-storey shearing-type numerical models and a 3-storey test frame structure are identified. Moreover, a "two-stage"method which predicts probable damaged areas first through modal sensitivity analysis and then calculates exact damages by optimising sensitivity equations with genetic algorithms is proposed. Meanwhile, it is used in the 20-storey numerical model and the 3-storey test structure. Finally, results obtained by genetic algorithms are compared with those by the least-square method. Convergence properties, genetic parameters and some strategies of genetic algorithms are discussed. Disadvantages of this proposed method are pointed out and the development of the future research on structural damage identification is also discussed. Numerical and experimental results show that the proposed method is successful to identify complex damages in shearing-type structures. Compared with the least-squared method, genetic algorithm owns better robustness and global search capacities. Furthermore, because it needs not structural parameters, FE models and bulk of measured data, this method is more applicable in practice. Meanwhile, for complicated multi-storey buildings, the proposed "two-stage"method may obtain satisfied results, and also can increase the degree of accuracy, as well as largely save computing resources.